Wind Directions Retrieval from SAR Image Combining the Spatial and Spectral Domains: A Novel Approach Using a Modified Circle Median Filter

Abstract

Sea surface wind is of great importance for weather forecasts, oil spill monitoring, etc. Currently, sea surface wind by using reversion techniques from synthetic aperture radar (SAR) images has become emerging means to get more fine-grained wind products. In general case, the retrieval of wind speeds from synthetic aperture radar (SAR) images requires the input of wind directions. There are two conventional methods of retrieving wind directions from SAR images, namely, two-dimensional fast Fourier transform (2D FFT) method in the spectral domain and local gradients (LG) method in the spatial domain. However, the spectral method works fine only on open oceans and large image areas, such as 20 × 20km and the spatial method is likely aligned with wind streak. To improve the accuracy and resolution of 2D FFT-retrieved wind directions, a new approach by combining the wind directions retrieved from 2D FFT and LG with a modified cycle median filter (CMF) is proposed in this paper. In our experiment, the proposed method is validated at different spatial resolutions, and the resolution of FFT-retrieved wind directions can be improved to 3km × 3km.